You're on the subway and suddenly remember you need to ask an AI about a code issue. Pull out your phone, open a terminal? Not practical. But what if you could chat with Claude Code directly through Feishu or DingTalk?
cc-connect does exactly that.
9,359 stars on GitHub, 863 forks, up 1,408 stars this week. Written in Go, it bridges locally-running AI coding agents to virtually every major messaging platform — Feishu/Lark, DingTalk, Slack, Telegram, Discord, LINE, and WeChat Work.
And most platforms don't require a public IP.
What problem it solves
Chinese developers face a unique challenge: we communicate primarily through Feishu and DingTalk, not Slack or Discord. But most AI agent tools (Claude Code, Cursor, Codex) are CLI-only — they only work in a local terminal.
This means once you step away from your computer, you lose access to your AI coding assistant.
cc-connect's approach is straightforward: run a local bridge service that forwards messages from chat platforms to your local AI agent, then sends the agent's responses back. The AI agent runs on your own machine throughout — your code never touches a third-party server.
Key technical details
No public IP required. This is the most practical feature. Feishu, DingTalk, and Telegram all support webhook push or long-polling modes, so your local service can continuously receive messages without exposing a public endpoint. cc-connect leverages these mechanisms.
Multi-agent support. Claude Code, Cursor, Gemini CLI, Codex — you're not limited to just one. Different chat groups can map to different agents based on task type.
Multi-platform in parallel. A single cc-connect instance can simultaneously connect to Feishu groups, DingTalk groups, and Telegram chats. Messages from each platform are handled independently.
Session isolation. Each chat channel gets its own independent agent session. The project context you discuss in Feishu group A won't leak into group B.
Who should use it
Remote collaboration. Team members who can't set up a local dev environment? They can just @cc-connect in a Feishu group to get AI help with code.
Fragmented time utilization. A few minutes waiting for an elevator — open Feishu on your phone and ask your AI coding assistant, no terminal needed.
Ops monitoring. Connect cc-connect to Claude Code running on a server, and you can remotely run diagnostics or check logs through a DingTalk group.
Caveats
Security. While code doesn't pass through third parties, the messaging platform itself sees your messages. For sensitive code, use private chats rather than group chats, or run on a self-hosted Feishu/DingTalk instance.
Performance. The AI agent runs on your local machine, so response speed depends on your hardware and network. Large project contexts may have slower initial responses.
Setup complexity. Different platforms have different接入 methods. Feishu and DingTalk require creating apps and configuring webhooks. The docs are detailed but first-time setup takes some time.
My take
cc-connect fills a gap: mobile access to AI coding tools.
In China's ecosystem, Feishu and DingTalk are the de facto work communication platforms. cc-connect embeds AI coding capability into existing workflows instead of demanding everyone learn a new tool. This "go where you already are" approach is more pragmatic than "come use a new product."
9,359 stars for this type of project is meaningful — the demand is real. If your team already uses Claude Code or Cursor, cc-connect is worth a try, especially for Feishu/DingTalk users.
Sources:
- GitHub - chenhg5/cc-connect (repository analysis)
- GitHub Trending Go Weekly (t热度追踪)
- Project README and documentation